File make-ha-ocf-reliable.patch of Package rabbitmq-server

From 1bf543bb7b3f36cf4240f8ea57cc34a825dc48bf Mon Sep 17 00:00:00 2001
From: Vincent Untz <vuntz@suse.com>
Date: Wed, 22 Nov 2017 11:41:49 +0100
Subject: [PATCH] Make OCF RA more reliable for old versions of rabbitmq

---
 scripts/rabbitmq-server-ha.ocf | 42 +++++++++++++++++++++++++++++-------------
 1 file changed, 29 insertions(+), 13 deletions(-)

diff --git a/scripts/rabbitmq-server-ha.ocf b/scripts/rabbitmq-server-ha.ocf
index f8e51ae..e1900b2 100755
--- a/scripts/rabbitmq-server-ha.ocf
+++ b/scripts/rabbitmq-server-ha.ocf
@@ -1413,6 +1413,12 @@ get_status() {
 
     body=$( ${COMMAND_TIMEOUT} ${OCF_RESKEY_ctl} eval 'rabbit_misc:which_applications().' 2>&1 )
     rc=$?
+    # Transient failure that happen from time to time with old rabbitmq
+    if echo "$body" | grep "TCP connection succeeded but Erlang distribution failed" > /dev/null 2>&1; then
+        ocf_log info "${LH} re-running command once after transient failure"
+        body=$( ${COMMAND_TIMEOUT} ${OCF_RESKEY_ctl} eval 'rabbit_misc:which_applications().' 2>&1 )
+        rc=$?
+    fi
 
     pgrep -f "beam.*${RABBITMQ_NODENAME}" > /dev/null
     beam_running=$?
@@ -1422,6 +1428,13 @@ get_status() {
         return $OCF_NOT_RUNNING
     # return a generic error, if there were errors and beam is found running
     elif [ $rc -ne 0 ] ; then
+        # The call is not fully reliable with an old version of rabbitmq, so deal with it
+        if [ $rc -eq 2 ] ; then
+            if echo "$body" | grep "node rabbit@.* up, 'rabbit' application running" > /dev/null 2>&1; then
+                ocf_log info "${LH} temporary failure, but diagnostics tell things are okay"
+                return $OCF_SUCCESS
+            fi
+        fi
         ocf_log info "${LH} found the beam process running but failed with code ${rc}. Command output: ${body}"
         return $OCF_ERR_GENERIC
     fi
@@ -1628,11 +1641,12 @@ get_monitor() {
         return $rc
     fi
 
-    # rc can be SUCCESS or RUNNING_MASTER, don't touch it unless there
-    # is some error uncovered by node_health_check
-    if ! node_health_check; then
-        rc=$OCF_ERR_GENERIC
-    fi
+    # Not reliable with an old version of rabbitmq, so commenting this out
+    ## rc can be SUCCESS or RUNNING_MASTER, don't touch it unless there
+    ## is some error uncovered by node_health_check
+    #if ! node_health_check; then
+    #    rc=$OCF_ERR_GENERIC
+    #fi
 
     if [ $rc -eq $OCF_RUNNING_MASTER ] ; then
         # If we are the master and healthy, perform various
@@ -1640,7 +1654,8 @@ get_monitor() {
         # Order a member to restart if something fishy happens with it.
         # All cross-node checks MUST happen only here.
 
-        partitions_report="$(partitions_report)"
+        # Not reliable with an old version of rabbitmq, so commenting this out
+        #partitions_report="$(partitions_report)"
 
         for node in $(get_alive_pacemaker_nodes_but $THIS_PCMK_NODE); do
             # Restart node if we don't consider ourselves clustered with it
@@ -1650,13 +1665,14 @@ get_monitor() {
                 continue
             fi
 
-            # Restart node if it has any unresolved partitions
-            node_partitions=$(grep_partitions_report $node "$partitions_report")
-            if [ ! -z "$node_partitions" ]; then
-                ocf_log warn "${LH} Node $node thinks that it is partitoned with $node_partitions"
-                order_node_restart "$node"
-                continue
-            fi
+            # Not reliable with an old version of rabbitmq, so commenting this out
+            ## Restart node if it has any unresolved partitions
+            #node_partitions=$(grep_partitions_report $node "$partitions_report")
+            #if [ ! -z "$node_partitions" ]; then
+            #    ocf_log warn "${LH} Node $node thinks that it is partitoned with $node_partitions"
+            #    order_node_restart "$node"
+            #    continue
+            #fi
         done
     fi
 
-- 
2.15.1